library.bib is in UTF8 BibLaTeX format, not bibtex format. BibLaTeX can be considered the successor to bibtex.

keywords.txt lists most of the keywords that I use to categorise entries. The list in keywords.txt is not exhaustive.

The files in python/ are IPython Notebook scripts that I have used for processing my .bib files using the excellent bibtexparser package.

Please feel free to re-use any of this content. Also, please do submit an issue or a pull request if you find any errors!!

• Energy disaggregation (aka NILM, NIALM, NALM)
• Behavioural / psychological studies surrounding NILM
• Climate science
• Climate change legislation
• Machine learning, especially deep learning
• Neuroscience

I try to use the capitalisation scheme used in the PDF version of the paper. Note that BibLaTeX respects the capitalisation in the title field without having to use double curly brackets (which bibtex requires). Note also that some publishers use different capitalisation schemes in the HTML version (or database entries) and PDF versions of papers. Hence why I try to use the capitalisation used in the PDF version.

Perhaps slightly controversially, I remove the In Proceedings and just leave the name of the conference. I also convert, say, 'second conference in...' to '2nd conference in...'. I'm tempted to use superscripts...

For conferences, I use series for the standard acronym for the conference (without any numbers). e.g. booktitle={25th International Conference on Machine Learning}, series={ICML}. This usage is a little unusual and perhaps I should instead do booktitle={25th International Conference on Machine Learning (ICML)}.

I extract the publisher from the conference title and put it into the publisher field. e.g. I do:

booktitle = {2nd International Conference on Embedded Systems for
Energy-Efficient Built Environments},
publisher = {ACM},
series = {BuildSys},
location = {Seoul, South Korea}

I try to use the full author names.

If a paper has a DOI which resolves to an open access PDF then, in general, I won't include a URL or eprint.

• dataset is a single string or a comma separated list of strings of the short name of the dataset(s) used. e.g. dataset = {UK-DALE, REDD}
• sampleperiod is the sample period of the aggregate data in seconds. Alternatively, use:
• samplefrequency is the sample frequency in Hz of the aggregate data.

I try to use UTF8 characters instead of ugly LaTeX accent codes. e.g. I prefer Bergés over Berg\'{e}s.

I haven't tried this yet but some things that you'd definitely need to change:

• rename the field journaltitle to journal
• biblatex uses the date field in preference to the year and month field, and date ranges are specified like 2015-01-15/2015-01-20. I'm not sure if bibtex will be happy with that.
• wrap titles in double curly brackets to preserve capitalisation.
• I expect that, for conferences, the contents of the publisher and series fields will need to be moved into the booktitle field.